Distributed Sparse Parameter Estimation Over Binary Sensor Networks

With the rapid development of network technology,microelectronics technology and low-power wireless communication technology,wireless sensor networks have become a hot research field.The sensor network integrates sensor technology,computer technology,information processing technology and communication technology,and can synchronously monitor,sense and collect the physical information of various environments or monitoring objects in the network coverage area in real time,and process and transmit them in the military.The fields of national defense,industrial and agricultural control,environmental testing,and remote control of hazardous areas all have important scientific significance and broad application prospects.Wireless sensor network is a collection of sensor nodes that are dispersed in a certain spatial area.Parameter estimation is one of the important applications of wireless sensor networks.It establishes certain statistical mathematical models and uses various algorithms to obtain an estimate of a certain parameter from noise-contaminated signal measurements to determine the interdependencies between different physical quantities.Sensor nodes are one of the key technologies of wireless sensor networks due to their limited communication range and limited node energy.The use of certain data fusion technologies can effectively reduce the communication load in the network and improve the performance of the entire network.At present,data fusion technologies are mainly divided into centralized and distributed.In a centralized sensor network,all nodes transmit the collected signals to the fusion center for processing.The signal utilization is high but depends on the fusion center.The fusion center needs strong storage and processing capabilities.The distributed sensor network does not need a fusion center,but the data is fused by the mutual cooperation between the nodes.Since the central node is no longer overly dependent,the whole network has better robustness.On the other hand,in a sensor network,each sensor node typically has limited computing,communication,and storage capabilities,and the nodes used in the sensor network have different performance and cost.In recent years,the theory and application research based on binary sensor networks have received extensive attention.Each node in the network can only provide low-precision 1-bit measurement values,and sensors that can provide analog measurement values(infinite precision).Compared,the binary sensor has a lower cost of use.In addition,1-bit transmission can greatly reduce network communication load and extend the life cycle of network nodes.Based on the traditional sensor network parameter estimation,this paper studies the adaptive parameter estimation algorithm based on 1-bit measurement of sensor network.The main innovations and achievements of the thesis are as follows:(1)Based on the analysis of traditional sensor network based LMS parameter estimation algorithm,this paper considers a parameter estimation problem using low-cost binary sensor network,and deduces a 1-bit measurement LMS parameter estimation method,and proposes a 1-bit distributed LMS algorithm and 1-bit centralized LMS algorithm.(2)Considering that many systems are sparse in practical applications,making full use of this sparsity can effectively improve the accuracy and robustness of parameter estimation.Therefore,the paper introduces a sparse constraint term in the parameter estimation algorithm of one-bit measurement,and proposes a centralized sparse LMS algorithm and a distributed sparse LMS algorithm based on 1-bit measurement of sensor networks.Through simulation experiments and comparison with traditional algorithms,the results verify the effectiveness of the proposed algorithm.